Hello! I recently created a project using restaurant data from OpenStreetMap. I was curious how the data compares to Overture Places restaurant data, so I did some analysis myself of it. Thought I’d share here!
Thanks for carrying out this assessment and writing up your findings. We need more extrinsic analyses along these lines to gauge completeness parity. Some suggestions for further steps:
- Refine your OSM confidence metric by ignoring edits by bots such as woodpeck_fixbot and b-jazz-bot (which doesn’t distinguish domain squatters as it touches
website
tags) - Repurpose the heuristics from intrinsic quality analyses as confidence factors
- Sample enough POIs to compare parity levels across geographies or population densities
- Compare OSM and Overture coverage of a wider variety of POI types, especially those that Google Maps may not cover as well
- Use another third-party data source as a baseline, such as AllThePlaces, Yelp, or Foursquare
I have a related question about comparing the two:
Is anyone aware of any work being done to map OSM tags to the OvertureMaps schema ( Reference | Overture Maps Documentation specifically for Places schema/docs/schema/concepts/by-theme/places/overture_categories.csv at main · OvertureMaps/schema · GitHub )?
There’s a big correspondence table on the wiki. Mostly it shows where we need to come up with new tags for common real-world phenomena.
Well, maybe. The Overture list looks like someone looked at a pile of advertising copy and tried to classify it, rather than being familiar with that thing in the real world.
Ad an example, I currently create maps which show about 9 different attributes for pubs based on OSM data. Almost none of those are catered for by Overture’s categorisation. Of the ones that sort-of are (“food”, “beer_garden”), there’s some doubt that Overture is even tracking these as attributes rather than POIs in their own right.
Some other categories, e.g. "Indo-Chinese Restaurant, might want a bit of checking to see if it’s really the same as “Indonesian”, or something else - how things are labelled varies hugely around the world.
Of course, I’m just going by the OSM wiki page - it may be hugely oversimplifying things, and missing a lot of context.
Yes, I agree that a lot of the contributions to that page have been questionable. It’s a wiki, edits welcome.
I’ve been maintaining similar pages for the North American POI classification standard. Occasionally someone will come by and try to fill in the blanks but fail to capture the essence of the distinctions being made. I see some of that happening in the Overture table too.